Dimensions of AI de-anthropomorphization based on impression integration theory

As artificial intelligence (AI) becomes increasingly anthropomorphized and integrated into the cognitive, emotional, and social domains of human life, a paradox emerges: despite its human-like design and interactional capabilities, users often deny its human-like qualities, a process we refer to as AI de-anthropomorphization. Drawing on impression integration theory, we conceptualize AI de-anthropomorphization along three core dimensions, cognitive, emotional, and relational, each comprising three specific subdimensions. Based on survey data, we identify a nine-factor structure that captures AI de-anthropomorphization, reflected in users' perceptions of AI as lacking rational thought, emotional depth, and social understanding. We further show that de-anthropomorphization is triggered by AI identity recognition and, in turn, shapes users’ attitudes toward AI. Theoretically, this study positions AI de-anthropomorphization as a central lens for understanding human–AI distinctions in contexts of increasing anthropomorphism. By showing that AI can be perceived as human-like and subsequently de-anthropomorphized, this research offers a novel typology and lays the foundation for future work on how such perceptions shape user behavior and evaluations in AI-related contexts.